← Back to Blog
OPINION·March 2026

Why Brands Need a Third Way Between Search Bars and AI Agents

The old way of finding products is broken. The new way threatens to break the brands that sell them. There's a better path.

Two things are simultaneously true in e-commerce right now, and they're pulling the industry apart.

The first: keyword search is failing. When 67% of shoppers abandon because they can't find what they want, and $350 billion is lost annually to poor product discovery, it's not a user experience problem. It's a structural failure. The search bar was designed for a world where people know exactly what they want and can describe it in two or three words. That world never really existed, and it certainly doesn't exist now.

Try typing "off-duty model vibes under $80" into any major retailer's search bar. You'll get a page of irrelevant results — or nothing at all. That's not a query problem. It's a technology problem. People think about products in terms of aesthetics, feelings, contexts, and constraints. Search bars understand keywords.

The second truth: the proposed replacement — autonomous AI agents that shop on your behalf — is creating a panic among the brands and platforms that depend on owning the customer relationship. And for good reason.

The operator's dilemma

If you're a VP of Digital Experience at a major retailer (as one of the voices in this conversation is, having built ML-based personalization at Walmart Labs and now running digital CX at one of America's largest grocers), you live in a very specific tension.

You know your search bar is inadequate. You know customers are already using ChatGPT and other AI tools to start their shopping journeys. The data is hard to ignore — AI referral traffic to major retailers has surged, with some seeing 15-20% of their referral clicks coming from ChatGPT alone.

But you also know that if an AI agent owns the full shopping experience — discovery through checkout — your brand becomes a fulfillment endpoint. An interchangeable product catalog inside someone else's app. Your retail media business (which might be your fastest-growing revenue stream) evaporates, because the AI agent bypasses all the advertising surfaces you've built.

This is why Amazon is blocking AI crawlers. It's why Walmart pulled out of OpenAI's Instant Checkout and embedded its own agent instead. It's why industry analysts are calling agentic commerce a "mirage" — not because the technology doesn't work, but because the economics don't support it.

The operators who actually run these businesses understand something the hype cycle misses: you can't build a sustainable commerce model on top of platforms that have every incentive to block you.

What each side gets wrong

The status quo defenders (keep the search bar, add some AI polish) are fighting a losing battle. Consumers have tasted conversational, contextual product discovery and they're not going back. A chatbot layered on top of the same keyword search index doesn't solve the underlying problem — it just puts a conversational interface on broken infrastructure.

The agentic maximalists (AI agents buy everything autonomously) ignore the inconvenient truth that platforms won't cooperate, consumers aren't ready, and brands have no incentive to participate in their own commoditization. As one analyst put it: if you believe in agentic commerce, use it for your next five purchases — including a $100+ purchase or a gift for someone you care about. Most people wouldn't.

The "add a chatbot" middle (slap a conversational UI on your existing site) mistakes the interface for the intelligence. A chatbot that queries the same keyword index with slightly better natural language processing doesn't solve the fundamental problem of understanding visual style, subjective taste, and evolving preferences.

All three approaches share the same flaw: they treat discovery and conversion as a single, monolithic experience that one party needs to own end-to-end.

The case for separation

What if discovery and conversion are actually two different problems that require two different architectures?

Discovery is about understanding. It requires:

  • Visual intelligence (understanding that a "boho chic midi dress" looks like a specific set of silhouettes, fabrics, and patterns)
  • Semantic understanding (parsing subjective, compound queries that keyword search can't handle)
  • Preference learning (adapting in real-time as the customer reacts to options)
  • Constraint handling (tracking budgets, sizes, exclusions, and dealbreakers)
  • Cold-start capability (working for a first-time anonymous visitor with no history)

This is where AI genuinely excels — and where search bars genuinely fail.

Conversion is about trust and execution. It requires:

  • Brand authority (the customer trusts this retailer to deliver quality)
  • Deep personalization (loyalty programs, purchase history, saved preferences)
  • Expert guidance (a knowledgeable associate, a detailed product page, size guides)
  • Transaction reliability (secure checkout, fulfillment, returns)
  • Relationship continuity (the customer comes back because they trust the brand)

This is where brands genuinely excel — and where AI agents genuinely struggle (as Walmart's Instant Checkout experiment demonstrated).

The insight is that these two capabilities are complementary, not competitive. The best shopping experience gives each to the party that does it best, with a clean handover in between.

The context handover

Here's what this looks like in practice:

A customer arrives — no login, no cookies, no purchase history. The AI engine starts a 60-second visual conversation. It shows products. The customer reacts. The AI learns their taste, narrows preferences, and surfaces constraints they might not have articulated ("I like this but shorter," "not that shade of blue," "under $120").

At the end of this process, the AI has built a structured context package:

  • Intent: What the customer is looking for, in semantic terms, not keyword terms
  • Preferences: Visual style, price range, specific attributes they gravitate toward
  • Constraints: Budget, sizing, timing, dealbreakers
  • What's been ruled out: Everything they've seen and rejected, and why

This context is then handed to the brand. Not as a cookie. Not as a retargeting pixel. As a structured, rich understanding of a qualified customer.

The brand receives this customer and can immediately offer deep personalization — because they know exactly what this person wants, what they've already considered, and what won't work. The journey doesn't reset. The customer doesn't have to re-explain themselves. The brand can focus on what it does best: guiding the final decision with expertise, loyalty incentives, and the trust it's built over years.

This is the "third way." Not a search bar. Not an autonomous agent. A contextual handover that respects what AI is good at, what brands are good at, and what customers actually want.

Why now?

Three forces are converging to make this model not just viable, but necessary:

Consumer behavior has shifted. Nearly 60% of US consumers have used AI tools for shopping help. They're not going back to typing "blue dress" into a search box. But they're also not ready to let an AI agent buy things for them autonomously. They want AI-assisted discovery with human-controlled purchase.

Platform economics are crystallizing. Amazon blocks agents. Walmart owns its agent. Target is building its own AI shopping app. The message from every major retailer is the same: we want AI-powered customer acquisition, but we will not let go of the customer relationship. Any AI commerce model that depends on platforms cooperating against their own interests is dead on arrival.

The technology is ready. Multi-modal AI that understands images, text, and behavior simultaneously can now deliver genuine taste learning — not just keyword matching with a conversational wrapper. Zero cold-start personalization from the first interaction is possible. The discovery side of the equation has caught up to the aspiration.

The brands that act on this convergence now — that integrate AI-powered discovery into their customer acquisition funnel while retaining ownership of the conversion — will have a structural advantage over those still debating whether to add a chatbot to their search bar.

The search bar is dying. Autonomous agents are a mirage. The context handover is the architecture that aligns incentives for everyone: AI providers, brands, and customers.

The journey should evolve, not reset. And the time to build for that future is now.

This post draws on analysis from Eric Seufert (MobileDevMemo), Jatin Pahuja (retail CX leader), Chris Sheldon (retail media), Juozas Kaziukėnas (e-commerce analyst), and industry reporting from Modern Retail, Wired, and Adweek.